AI litigation continues to define the future of intellectual property law

A growing body of litigation across the US and Europe is beginning to establish the legal boundaries governing the use of intellectual property in artificial intelligence (AI) development.

Recent decisions involving AI training data, copyright, trademarks, and digital content are providing greater clarity for rights holders and technology companies alike, while highlighting the need for new legal frameworks as generative AI continues to evolve.

Anthropic settlement marks a landmark copyright development

One of the most significant developments is the approval of a USD 1.5 billion class-action settlement between AI developer Anthropic and authors and publishers whose books were allegedly used without authorization to train the company’s Claude large language model.

Approved by the US District Court for the Northern District of California, the settlement is believed to be the largest copyright settlement in US history. It covers more than 14,000 titles published by Bloomsbury Publishing, with eligible authors and the publisher expected to share compensation of approximately USD 3,000 per title after legal costs.

The agreement followed a landmark ruling by Judge William Alsup, who concluded that training AI models on copyrighted books constituted fair use under US copyright law. However, the court also held that Anthropic infringed copyright by creating and maintaining a central digital library containing more than seven million books, distinguishing the use of works for AI training from the unauthorized storage of copyrighted material.

The decision is expected to influence future licensing negotiations and litigation involving AI developers and content owners.

Getty Images ruling clarifies the limits of copyright and trademark claims

In the UK, the High Court of England and Wales largely ruled in favor of Stability AI in its closely watched dispute with Getty Images, delivering one of the first detailed judicial analyses of copyright and trademark issues arising from generative AI.

Getty Images abandoned its primary copyright infringement claims during the proceedings after it was unable to establish that the training of Stable Diffusion had taken place within the jurisdiction of the UK courts.

The High Court also dismissed Getty’s secondary copyright infringement claims, finding that the Stable Diffusion model weights do not store or reproduce copies of the original photographs. Instead, the court concluded that the model functions by learning statistical patterns from training data rather than retaining or distributing the underlying copyrighted images.

The court did, however, identify limited trademark infringement relating to early versions of Stable Diffusion. Some AI-generated images were found to contain distorted Getty Images or iStock watermarks, creating a limited infringement of Getty’s registered trademarks.

Importantly, the court described these watermark issues as historic and confined to earlier versions of the model, with no evidence that newer releases, including Stable Diffusion XL, continued to generate infringing outputs.

The ruling provides valuable guidance for developers and rights holders alike, suggesting that while AI training itself may avoid copyright liability under certain circumstances, AI-generated outputs remain capable of infringing trademark rights where protected branding appears in generated content.

Courts continue to examine AI training and fair use

Elsewhere, courts continue to test the application of existing copyright principles to AI development.

In Thomson Reuters v. Ross Intelligence, a US federal court held that the use of Westlaw’s editorial headnotes to train an AI-powered legal research platform did not qualify as fair use, finding that the copying served a commercial purpose and reproduced protected editorial expression.

Meanwhile, courts in Europe are also being asked to determine whether the reproduction of protected works during AI training falls within existing copyright exceptions, with several proceedings expected to shape the future regulatory landscape for generative AI.

These cases demonstrate that while some judicial decisions have recognized AI training as potentially transformative, the commercial acquisition, reproduction, and exploitation of copyrighted content continue to receive close scrutiny.

Deepfakes expose emerging gaps in IP protection

Beyond copyright and trademarks, AI-generated deepfakes and digital replicas continue to expose areas where existing intellectual property laws offer only limited protection.

Legal commentators have increasingly argued that jurisdictions such as the UK should consider introducing standalone personality or image rights, giving individuals greater control over the commercial use of their likeness, voice, and identity in AI-generated content.

Current legal remedies often require claims based on copyright, passing off, privacy, or data protection, leaving uncertainty where individuals seek to challenge unauthorized AI-generated representations that do not neatly fit within existing causes of action.

A rapidly evolving legal landscape

As AI technologies become more sophisticated, courts are beginning to establish clearer principles governing the relationship between artificial intelligence and intellectual property.

The Anthropic settlement demonstrates that AI developers may face substantial financial exposure where copyrighted works are acquired or stored without authorization, while the Getty Images judgment illustrates that successful copyright claims against AI developers may depend on jurisdictional and technical considerations. At the same time, the High Court’s findings on trademark infringement confirm that AI-generated outputs themselves can create legal risks where protected brands or logos appear in generated content.

For intellectual property practitioners, these decisions represent important milestones in the development of an increasingly sophisticated legal framework that will shape AI innovation, licensing strategies, and enforcement for years to come.

Interested in this topic? Check out our conversation with Dr Hayleigh Bosher about whether UK IP law is ready for the creative revolution here. 

← Back to News